Where the assumption came from
Platform reporting was taken at face value and blended ROAS was the single number in the weekly review. Every incremental pound looked profitable, so spend rose steadily with no test to confirm the causal link.
← Engagement stories
Ecommerce & retailHealth & supplements3-week diagnostic, incrementality testing over 6 weeks
The channel that looked free
Reported ROAS answers a question about attribution. It does not answer whether the spend caused the sale.
- Scale
- £8m–£12m online revenue
- Engagement
- Growth Diagnostic
- Client
- Anonymised by default
Quantified movement
Each figure is a measured band against the pre-engagement baseline stated beside it. Ranges are used because the underlying numbers are real and anonymised, not rounded into a single headline claim.
Measured incrementality of lower-funnel spend
38% to 52% of reported revenue
Baseline: Platform-reported conversions on the largest line item
Six-week geo holdout
Lower-funnel budget reallocated
≈24% of monthly media
Baseline: Baseline monthly paid media budget
Within one quarter
Contribution margin
+6% to +9%
Baseline: Baseline margin at flat total revenue
Quarter following reallocation
- UK region
- North West England
- Company stage
- Scaling (£5m–£15m)
- Constraint found
- Measurement and incrementality
Presented as
"Our reported blended ROAS is healthy, so we should spend more."
The real constraint
Measurement integrity. Branded search and lower-funnel retargeting were being credited with demand generated elsewhere, so a meaningful part of the media budget was buying customers the business already had.
What the diagnostic looked at
Attribution model overlap, share of conversions credited to lower-funnel activity, the relationship between spend and total revenue over time, and the availability of clean holdout capability in each platform.
What it found
Overlapping credit across platforms exceeded total actual orders in several weeks — a straightforward sign of double counting. A geo holdout on the largest lower-funnel line item showed materially less incremental revenue than reported. Upper-funnel activity, which reported poorly, was carrying more of the demand than the model credited it with.
What changed
A standing incrementality testing programme was set up with a repeatable holdout protocol, the weekly review moved to contribution margin against total revenue rather than blended ROAS, and budget was rebalanced toward the activity that testing showed was generating demand rather than harvesting it.
Would the same diagnosis hold in your business?
The Growth Economics Snapshot takes a few minutes and shows where your biggest commercial lever sits before you commit to anything.